Back to Toolkit
PhotogrammetryFree
Meshroom
Free open-source photogrammetry pipeline built on the AliceVision framework. Reconstruct detailed 3D meshes from sets of photographs using structure-from-motion and multi-view stereo algorithms. Useful for scanning real-world objects, environments, and textures as reference or base meshes for game art. Supports GPU-accelerated processing and exports to OBJ format for cleanup in Blender or ZBrush.
Best For
Converting real-world photo sets into dense 3D meshes via photogrammetry, providing high-quality reference geometry and texture data for game art retopology and scan-based asset creation.
Primary Use Cases
- - Photographing a real rock or brick wall to generate a photogrammetry mesh used as reference for sculpting a matching hero prop in ZBrush
- - Capturing a historical artifact or costume piece to extract real-world texture detail for baking into a game-ready PBR material in Substance Painter
- - Generating a dense terrain scan mesh from drone photographs to use as height reference for sculpting a game environment's landscape layer
- - Creating a photo-real foliage texture sheet by scanning real leaves and extracting albedo, normal, and translucency maps from the photogrammetry output
- - Building a high-poly human face scan from a structured-light photo session to use as a sculpting base for a realistic game character head
Pros & Cons for Game Art
Pros
- + Completely free and open-source — delivers professional photogrammetry output with zero licensing cost for students
- + Node-based pipeline graph is inspectable and modifiable, giving technical artists visibility into each reconstruction step
- + Outputs dense meshes with baked texture that provide higher-fidelity sculpting reference than any hand-painted or procedural alternative
- + Integrates naturally with ZBrush and Blender retopology workflows — the high-poly scan drives the detail baking stage directly
- + Large active community providing preset pipelines for common scan scenarios (faces, architecture, vegetation)
Cons
- − Requires Nvidia CUDA GPU — artists on AMD hardware must rely on slow CPU fallback or use a cloud GPU service, adding cost and friction
- − Processing time for a 200-photo set can exceed 4–8 hours on mid-range hardware, making rapid iteration impractical for a student with one machine
- − Reconstruction fails or produces holes on reflective, transparent, or featureless surfaces — glass, metal, painted walls all require special capture techniques
- − Output mesh density (often 5–50 million polygons) demands a high-RAM workstation for Blender import; 16 GB RAM frequently causes crashes on large scans
Suggested Learning Path
- 1.Photograph a simple, high-texture object (rough stone, worn boot) from 30–50 overlapping angles and run a full Meshroom pipeline to understand what each node does
- 2.Compare the output mesh quality at different FeatureExtraction and DepthMap settings — learn which parameters trade processing time for reconstruction density
- 3.Import a Meshroom OBJ into Blender and practice retopologizing the dense scan mesh onto a clean, game-ready base mesh using Blender's shrinkwrap modifier
- 4.Bake the scan mesh's texture detail (color, normal, AO) onto the clean retopology mesh in Marmoset Toolbag for a game-ready texture set
- 5.Photograph a set of tileable surface materials (stone, wood, concrete) and use the Meshroom textures as a starting point for Substance Painter material creation
Quick Tip
Shoot in overcast daylight — clouds act as a giant diffuse light box, eliminating harsh shadows that confuse Meshroom's depth estimation and produce holes in the reconstructed mesh.
Setup Notes
- - Meshroom requires an Nvidia CUDA GPU for its DepthMap node — without it, reconstruction falls back to CPU and takes 10–20x longer on typical photo sets
- - Set the FeatureExtraction preset to "Ultra" for complex objects and "Normal" for flat surfaces — over-extracting features on simple geometry wastes hours with no quality benefit
- - Use RAW or high-quality JPEG photos with consistent exposure; Meshroom's color consistency algorithms are sensitive to exposure variation between shots that causes patchy texture output
- - Export the final mesh as OBJ with the embedded UDIM texture rather than the default EXR texture set — OBJ is more universally compatible with Blender and ZBrush import workflows
How N-hance Uses It
N-hance introduces Meshroom in the 3D Character Artist course as a reference capture tool, teaching students to photograph real-world surfaces and clothing to extract texture and form data that informs ZBrush sculpting rather than relying solely on photo reference boards.